[zargon]

Open source is about freedom to modify the product. So in the context of an LLM, the source code is the data and the code that processes the data during *training* (not only inference), as that is what generates the weights.